Te dejo una opción con PROMPT para el video sin la voz que me pareció más práctica (dado que tuviste tan malos resultados antes) y luego tres PROMPTs alternativos junto con una "casi" pseudo recomendación para crear un mega master cuaderno de NotebookLM para crear prompts en base a fuentes oficiales junto con el PROMPT para forzar la salida de las respuestas en el formato deseado.
SOLUCION PRÁCTICA:
Creo que podrías echar mano de la edición con alguna herramienta tipo CapCut (o Clipchamp en Windows) o la que sea que te permita trabajar video por un lado y sonido en "capas" dentro de lo que se llama "linea de tiempo" en los programas de edición.
Porque podrías hacerlos que estén reunidos y que de pronto miren hacia el punto desde donde se supone que vendría la voz (sin mencionárselo a la IA para que no confunda un video en silencio con algo en donde debe aparecer "voz").
.
Ahí podrías poner la voz en el momento en que todos miran atentos hacia el punto desde donde se supone que se oye el sonido de la voz. y el audio, lo colocarías justo en el momento en que todos giran sus cabezas hacia ese punto determinado.
Sólo te lo digo para ahorrar tiempo y gasto innnecesario de tokens en tu IA de generación de video.
Para el video, te sugiero este PROMPT
sin la voz en off:
[IMAGE REFERENCE]: Use the exact visual identity, faces, and clothing of the people provided in the reference image. Maintain strict character consistency.
[SUBJECT & POSE]: A group of people (extracted from the reference photo) standing in a frame. Their mouths remain completely closed, tightly shut, and motionless throughout the video. ABSOLUTELY NO talking, NO lip movements, NO speech animation.
[ACTION & TIMING SEQUENCE]: - 0 to 3 seconds: The people are static, facing forward toward the camera with neutral, calm expressions. - At second 3: In a sudden, perfectly synchronized motion, ALL people simultaneously turn their eyes and heads to look directly at the top-left corner of the video frame. - 3 to 6 seconds: Everyone freezes their gaze on the top-left corner, holding their attention fixed on that specific spot without moving their lips or speaking.
[STYLE & CINEMATICS]: Photorealistic, smooth 4K movement, subtle natural lighting, static mouths, zero facial distortion.
-------------------------------------
PROMPT alternativo
con la voz en off:
[IMAGE REFERENCE]: Use the exact visual identity, faces, and clothing of the people provided in the reference image. Maintain strict character consistency.
[SUBJECT & POSE]: A group of people (extracted from the reference photo) standing in a frame. Their mouths remain completely closed, tightly shut, and motionless throughout the entire video. ABSOLUTELY NO talking, NO lip movements, NO speech animation from the subjects.
[AUDIO / VOICE OVER]: An off-camera voice (originating from off-set, near the top-left corner outside the frame) speaks in clear, natural Spanish: "¿Hola, cómo están?".
[ACTION & TIMING SEQUENCE]:
- 0 to 2 seconds: The people stand still with neutral, calm expressions, looking straight ahead toward the camera. Mouths tightly shut.
- 2 to 4 seconds: The off-camera voice-over speaks from the top-left direction outside the frame: "¿Hola, cómo están?". The subjects listen while remaining motionless.
- At second 4 (Immediately after the voice ends): In a sudden, perfectly synchronized reaction, ALL people simultaneously turn their eyes and heads toward the top-left corner of the frame where the voice originated.
- 4 to 6 seconds: Everyone freezes their gaze on the top-left corner, holding their attention fixed on that exact spot, with their mouths remaining completely motionless.
[STYLE & CINEMATICS]: Photorealistic, smooth 4K movement, subtle spatial audio from top-left, static mouths, zero facial distortion.
---------------------------------
Y TE DEJO DOS PROMPTS MAS! Uno SIN la voz en OFF y otro CON la voz en OFF:
Los Prompts sugeridos los he creado con NotebookLM en base a una base de conocimiento de las diferentes bibliografías de las diferentes IA de generación de imagen y video, configurando el chat del cuaderno con el siguiente PROMPT:
Actúa como un ingeniero de prompts. Tu objetivo es generar prompts para generación de contenido. Siempre consulta las fuentes antes de responder.
Ten en cuenta cuando te pidan un prompt:
1) Utilizar la sintaxis de parámetros exacta que aparece en las fuentes de acuerdo a la herramienta de IA que se utilizará (por ejemplo, códigos de cámara específicos, etiquetas de estilo o prompts negativos).
2) No inventes comandos que no existan en la documentación. Da aviso al usuario para que cargue documentación relacionada si no cuentas con fuentes suficientes.
3) Escribe el PROMPT íntegramente en inglés.
PROMPT SIN la voz en OFF:
```
Using the provided reference image containing the group of people as the foundational visual anchor:
[Cinematography]
Medium wide shot, static camera, eye-level framing, deep focus, steady composition.
[Subject]
The exact group of people extracted from the provided source photo. Their mouths remain firmly closed, completely silent, and motionless with no speaking, no lip movement, and no talking throughout the entire clip.
[Action]
[00:00-00:03] The people stand calmly in their original positions from the photo, remaining completely still with neutral facial expressions, eyes looking directly forward, mouths sealed closed.
[00:03-00:06] Simultaneously and in unison, all people slowly turn their heads and gazes toward the upper-left corner of the video frame. They freeze their gaze, staring intently and attentively at that exact upper-left spot until the end of the video, maintaining complete silence and closed lips.
[Context]
The original environment and background from the reference photo, maintaining full character identity, spatial coherence, and lighting consistency.
[Style & Ambiance]
Photorealistic, cinematic lighting, natural ambient shadows, atmospheric tension, photorealistic texture preservation.
Negative prompt: talking, speaking, lip movement, open mouth, mouth moving, vocalization, dialogue, facial distortion, morphing features, moving lips.
```
PROMPT CON la voz en OFF:
```
Using the provided reference image containing the group of people as the foundational visual anchor:
[Cinematography]
Medium wide shot, static camera at eye level, steady framing, deep focus.
[Subject]
The exact group of people extracted from the provided source photo. Their mouths remain firmly closed, motionless, and completely silent throughout the entire clip, with zero lip movement or speech from any person on camera.
[Action]
[00:00-00:02] The people stand in their original positions from the photo, remaining completely motionless with neutral expressions, eyes facing forward, mouths sealed closed.
[00:02-00:04] An off-screen voiceover speaks from outside the frame set. Off-screen voiceover says in Spanish: "¿Hola cómo están?".
[00:04-00:08] Immediately upon hearing the off-screen voice, all people simultaneously and in unison slowly turn their heads and gazes toward the upper-left corner of the video frame. They hold their gaze, staring attentively at that exact upper-left spot until the end of the clip, maintaining strictly sealed lips.
[Context]
The original environment and background setting from the reference photo, maintaining full character identity, spatial consistency, and lighting coherence.
[Style & Ambiance]
Photorealistic, cinematic natural lighting, atmospheric tension, photorealistic skin and fabric texture preservation.
Negative prompt: talking, speaking, lip movement, open mouth, moving lips, vocalizing, mouth animation.
Espero que te sirva !!!
PD:
Si necesitás que sus rostros o cabezas giren mirando hacia otra parte, deberías modificar la dirección en donde aparezcan cosas en el PROMPT como "top-left corner" ó "upper-left spot", etc.
IMG_0174.webp172,2 KB · 1 visualizaciones
